3.1 The clinical effectiveness data presented by the manufacturer were predominantly from 2 phase III open-label randomised controlled clinical trials. LUX-Lung 3 compared afatinib with cisplatin plus pemetrexed and LUX-Lung 6 compared afatinib with cisplatin plus gemcitabine. There was also a mixed treatment comparison that compared afatinib with erlotinib and gefitinib. LUX-Lung 3 was an international trial (ethnicity: 26% white, 72% Eastern Asian, 2.0% other) that compared afatinib (40 mg daily, n=230) with cisplatin plus pemetrexed (n=115) for patients with EGFR mutation-positive NSCLC. LUX-Lung 6 was conducted in China, Thailand and South Korea and compared afatinib (40 mg daily, n=242) with cisplatin plus gemcitabine (n=122) for patients with EGFR mutation-positive NSCLC. In both trials patients were included who had received no prior treatment with chemotherapy or EGFR-targeting drugs and adenocarcinoma was the predominant histology. The primary outcome of the clinical trials was progression-free survival, as assessed by central independent review by Response Evaluation Criteria in Solid Tumours (RECIST version 1.1). Secondary outcomes included objective response rate and overall survival.
3.2 In LUX-Lung 3, there was a statistically significant increase in median progression-free survival for afatinib compared with cisplatin plus pemetrexed combination chemotherapy (11.14 months compared with 6.90 months; a gain of 4.24 months) with a hazard ratio of 0.58 (95% confidence interval [CI] 0.43 to 0.78) when assessed by independent review. When the outcome was assessed by the trial investigator, there was a statistically significant increase in median progression-free survival for afatinib compared with combination chemotherapy (11.07 months compared with 6.70 months; a gain of 4.37 months) with a hazard ratio of 0.49 (95% CI 0.37 to 0.65).
3.3 In LUX-Lung 6, there was a statistically significant increase in median progression-free survival for afatinib compared with cisplatin plus gemcitabine combination chemotherapy (11.01 months compared with 5.59 months; a gain of 5.42 months) with a hazard ratio of 0.28 (95% CI 0.20 to 0.39) when assessed by independent review. When the outcome was assessed by the trial investigator, there was a statistically significant increase in median progression-free survival for afatinib compared with combination chemotherapy (13.73 months compared with 5.55 months; a gain of 8.18 months) with a hazard ratio of 0.26 (95% CI 0.19 to 0.36).
3.4 In LUX-Lung 3, overall survival data were not mature by the cut-off date (February 2012) for the primary analysis because 67 patients (29.1%) in the afatinib arm and 31 patients (27.0%) in the chemotherapy arm had died. The manufacturer presented the results of the updated analysis (using additional data after the February 2012 cut-off) and the results of an updated analysis submitted to the European Medicines Agency using data up to January 2013). The manufacturer stated that final analysis of overall survival will be performed when 209 patients have died. No statistically significant difference in overall survival was seen in LUX-Lung 3 or LUX-Lung 6 between afatinib and chemotherapy with hazard ratios of 1.12 (95% CI 0.73 to 1.72) and 0.95 (95% CI 0.68 to 1.33) respectively. Treatment crossover occurred in both LUX-Lung 3 (72%) and LUX-Lung 6 (80%) with most patients receiving at least 1 line of subsequent anticancer therapy after stopping the study drugs. Crossover was not accounted for when estimating overall survival. The manufacturer conducted subgroup analyses of LUX-Lung 3 and LUX-Lung 6 for pre-specified baseline characteristics such as gender, age, family origin and common EGFR mutations, which was consistent with the analysis in the intention-to-treat populations.
3.5 Health-related quality of life data from LUX-Lung 3 and LUX-Lung 6 were reported for the pre-specified NSCLC-related symptoms of cough, dyspnoea and pain, measured by the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC QLQ)-C30 and QLQ-LC13 questionnaires. More than 87% of patients completed the questionnaires. Afatinib was associated with a statistically significant improvement in breathing, non-specific pain and chest pain, fatigue and the time to deterioration in cough, dyspnoea and pain compared with chemotherapy (cisplatin plus either pemetrexed or gemcitabine). EQ-5D (UK and Belgium) and EQ-VAS data collected during the LUX-Lung 3 clinical trial reported no statistically significant difference in values between afatinib and chemotherapy with an absolute improvement in utility of 0.008 (UK) and 0.007 (Belgium).
3.6 Because there was no head-to-head randomised controlled trial comparing the effectiveness of afatinib with erlotinib or gefitinib for progression-free survival or overall survival, the manufacturer presented a mixed treatment comparison. This was based on a previous mixed treatment comparison conducted for Gefitinib for the first-line treatment of locally advanced or metastatic non-small-cell lung cancer (NICE technology appraisal guidance 192), which was adapted to include data on the effectiveness of afatinib based on the LUX-Lung 3 and 6 studies and erlotinib.The studies used to populate the mixed treatment comparison were identified through systematic review. The manufacturer identified 20 randomised controlled trials, 4 of which included gefitinib (first SIGNAL trial, IPASS trial, Mitsudomi 2010, Maemondo 2010) and 1 that included erlotinib (EURTAC trial). The population of the studies included in the mixed treatment comparison was people with locally advanced or metastatic non-small cell lung cancer. However, only 7 of the trials were carried out exclusively in people with EGFR-positive disease. The manufacturer specified that the EURTAC trial was of average quality and the first SIGNAL trial only included 42 patients with EGFR-positive disease. All the trials included in the mixed treatment comparison permitted crossover after disease progression. A fixed-effects model was used to assess progression-free survival and a random-effects model was used to assess overall survival. There was no testing of the proportional hazards assumption.
3.7 The results of the manufacturer's mixed treatment comparison showed that there was no statistically significant difference in progression-free survival or overall survival between afatinib and gefitinib or erlotinib. Afatinib had the highest probability (62.6%) of being the most effective treatment in terms of progression-free survival gain compared with all the comparator treatments including erlotinib (30.8%) and gefitinib (6.5%). Afatinib also had the highest probability (43%) of being the most effective treatment in terms of overall survival gain compared with all the comparator treatments including erlotinib (3%) and gefitinib (13%).
3.8 The resulting hazard ratios from the mixed treatment comparison for the difference in median progression-free survival for afatinib were:
0.36 (95% CI 0.25 to 0.52) compared with gemcitabine plus cisplatin
0.46 (95% CI 0.32 to 0.66) compared with pemetrexed plus cisplatin
0.78 (95% CI 0.47 to 1.20) compared with gefitinib
0.91 (95% CI 0.53 to 1.50) compared with erlotinib.
3.9 The resulting hazard ratios from the mixed treatment comparison for the difference in median overall survival for afatinib were
0.86 (95% CI 0.67 to 1.10) compared with gemcitabine plus cisplatin
0.99 (95% CI 0.78 to 1.27) compared with pemetrexed plus cisplatin
0.84 (95% CI 0.55 to 1.30) compared with gefitinib
0.80 (95% CI 0.56 to 1.14) compared with erlotinib.
3.10 The manufacturer submitted evidence from the non-placebo controlled LUX-Lung 2 trial, a phase II multicentre trial conducted in the USA and Taiwan, which evaluated the safety and efficacy of afatinib in patients with locally advanced or metastatic EGFR-positive NSCLC. The patients in the study were predominantly Asian. This trial evaluated the effectiveness of afatinib in patients who had not previously received chemotherapy (n=61) and patients whose disease had progressed after 1 previous chemotherapy treatment (n=68). There were 2 study arms, afatinib 40 mg and afatinib 50 mg. The primary outcome of LUX-Lung 2 was progression-free survival, as assessed by central independent review by RECIST version 1.1. Secondary outcomes included objective response rate and median overall survival.
3.11 In LUX-Lung 2, progression-free survival was shorter in patients who had received prior chemotherapy. The median progression-free survival was 11.9 months for patients who had not had chemotherapy before and 4.5 months for patients receiving the 40 mg dose of afatinib as a second-line treatment. LUX-Lung 2 reported shorter overall survival in patients who had prior chemotherapy compared with those who had not. The median overall survival was 23.1 months for patients who had not had chemotherapy and 14.6 months for patients receiving afatinib as a second-line treatment.
3.12 LUX-Lung 3 reported higher rates of diarrhoea, rash/acne, stomatitis/mucositis and paronychia compared with chemotherapy but less nausea, fatigue, vomiting, anaemia, leukopenia and neutropenia. The manufacturer compared the adverse reactions from the pivotal clinical trials for afatinib (LUX-Lung 3), gefitinib (IPASS) and erlotinib (EURTAC), which showed that afatinib is associated with more diarrhoea (95%) than gefitinib (47%) and erlotinib (57%), more rash/acne (89%) than gefitinib (66%) and erlotinib (80%), more stomatitis/mucositis (72%) than gefitinib (17%), but less reduced appetite (21%) than erlotinib (53%) and less fatigue (18%) than erlotinib (47%). Dose reductions were higher with afatinib (57%, LUX-Lung 3) compared with gefitinib (16.1%, IPASS) or erlotinib (21%, EURTAC).
3.13 The manufacturer presented a de novo disease-state cohort model consisting of 2 health states (progression-free and progressive disease) and death. The progression-free health state represented the period during which the patient's cancer did not worsen while receiving active treatment. The progressive disease health state represented the period that the cancer spread. The model allowed movement from the progression-free health state to the progressive-disease health state, or death state; or from the progressive-disease health state to the death state. The model had a lifetime horizon of 10 years and a cycle length of 1 month, with an NHS and personal social services perspective and 3.5% discounting for costs and health effects.
3.14 The manufacturer's model used the partitioned survival method to determine the proportion of patients in each health state, for each model cycle. Data from LUX-Lung 3 and LUX-Lung 6 were used to estimate progression-free survival and overall survival for afatinib in the model, but parametric survival models based on hazard ratios produced from the mixed treatment comparison were used to project progression-free survival and overall survival over the 10-year model time horizon. The Weibull method was used to extrapolate the trial data in the base-case model to estimate progression-free survival and overall survival. Sensitivity analyses were conducted that used 2 further types of parametric survival modelling of the clinical trial Kaplan-Meier data: exponential and Gompertz. The progression-free survival and overall survival estimates for people treated with erlotinib and gefitinib were estimated by applying the mixed treatment comparison hazard ratios to the survival estimates for people treated with afatinib. Progression-free survival and overall survival were incorporated into the cost-effectiveness model by using full parametric approximation or by using Kaplan-Meier data from the clinical trials extrapolated using parametric survival models.
3.15 Adverse reactions (diarrhoea, rash/acne, fatigue, anaemia and neutropenia) were applied in the model for the first year only, in both the progression-free and progressive disease health states. The type and frequency of adverse reactions was estimated from LUX-Lung 1 and LUX-Lung 3 for afatinib, and from the mixed treatment comparison for gefitinib and erlotinib.
3.16 In the base-case model the utility value used in the progression-free health state was 0.78 (from LUX-Lung 3) and utility values from the literature were used for the progressive disease health state (0.73 and 0.46 for second- and third-line treatment respectively). Alternative utility values derived from the literature for the progression-free health state were used in a sensitivity analysis. Utility values did not change over time.
3.17 To estimate the costs in the model, the manufacturer either used resource costs from the LUX-Lung 3 and 6 trials, or values from the literature. The resource costs associated with disease management (progression-free or progressed disease health states) and adverse reactions estimated from the LUX-Lung 3 and 6 trials included:
outpatient visits (GP, specialist, nurses, occupational therapist, physiotherapist)
outpatient interventions (CT scan, MRI scan, surgical procedure, ultrasound, X-ray, radiotherapy)
unscheduled hospitalisations (unscheduled hospital stay, intensive care unit visit, emergency room visit)
All other values were taken from the literature. The model assumed that treatment with afatinib, erlotinib or gefitinib continues until disease progression. Disease progression is typically assessed every 3 months by CT scan, and this cost was incorporated into the model. Afatinib, gefitinib, and erlotinib each have patient access schemes agreed with the Department of Health, which were accounted for in the analyses.
3.18 The deterministic pairwise results of the base-case analysis showed that afatinib was associated with an ICER of £10,076 per QALY gained (incremental costs £1723, incremental QALYs 0.171) compared with erlotinib and an ICER of £17,933 per QALY gained (incremental costs £3113, incremental QALYs 0.173) compared with gefitinib. The manufacturer stated that there was a 100% probability of afatinib being cost effective compared with erlotinib at £20,000 and £30,000 per QALY gained. Compared with gefitinib, there was a 72% and 81% probability of afatinib being cost effective at £20,000 and £30,000 per QALY gained respectively.
3.19 The manufacturer conducted one-way sensitivity analyses of the pairwise comparisons with gefitinib and erlotinib. The main drivers of cost effectiveness were: the mixed treatment comparison-based hazard ratios for progression-free and overall survival, the cost per month for the progression-free health state and the cost per month for the best supportive care period of the progressive disease health state. Overall, the ICERs estimated for the one-way sensitivity analyses ranged from £7135 to £54,800 per QALY gained for afatinib compared with gefitinib, and from −£10,302 to £34,970 per QALY gained for afatinib compared with erlotinib.
3.20 The manufacturer conducted some scenario analyses that varied the choice of second-line treatment (using pemetrexed rather than docetaxel as the second-line treatment), the duration of second-line treatment, the utility values in the progression-free health state and the studies used in the mixed treatment comparison. Using pemetrexed as second-line treatment had a minimal impact on the ICER. Applying a proportional duration of second-line treatment increased the ICER for afatinib compared with gefitinib to a maximum of £19,952 per QALY gained and £15,718 per QALY gained compared with erlotinib. Applying utility values derived from the literature for the progression-free health state also increased the ICER, most notably when afatinib was compared with gefitinib, which resulted in an ICER of £20,256 per QALY gained. For the comparison of afatinib with erlotinib, changing the utility values had a minimal impact on the ICER. Using only LUX-Lung 3 data in the mixed treatment comparison for afatinib (that is, excluding LUX-Lung 6, which was based in Asia) had the most impact on the ICER. It increased the ICER for afatinib compared with gefitinib to £24,339 per QALY gained, but had the opposite effect on the comparison with erlotinib in which afatinib dominated erlotinib (that is, was cheaper and more effective). When only LUX-Lung 3 data and data from the OPTIMAL trial of erlotinib comparing carboplatin plus gemcitabine were included, afatinib had an ICER of £15,257 per QALY gained when compared with gefitinib, and £13,013 per QALY gained when compared with erlotinib.
3.21 The ERG stated that the lack of a significant overall survival benefit with afatinib in the LUX-Lung 3 and 6 trials may have been masked by the high rates of crossover. The ERG considered Asian and non-Asian populations to be relevant subgroups. In response to the ERG request for clarification the manufacturer provided a subgroup analysis using updated data from LUX-Lung 3, which showed that Asian patients treated with chemotherapy may have a different progression-free survival and overall survival compared with non-Asian patients. The ERG undertook exploratory analyses using the manufacturer's data and noted that the mean expected post-progression survival was different for patients treated with afatinib in the Asian subgroup than in the non-Asian subgroup. The estimated mean progression-free survival in Asian patients was 19.5 months for afatinib and 8.7 months for pemetrexed plus cisplatin and in non-Asian patients was 14.8 months for afatinib and 4.7 months for pemetrexed plus cisplatin. The estimated mean overall survival in Asian patients was 37.3 months for both afatinib and pemetrexed plus cisplatin and in non-Asian patients was 31.4 months for afatinib and 25.3 months for pemetrexed plus cisplatin.
3.22 The ERG considered the population of the trials included in the mixed treatment comparison in light of evidence from the subgroup analysis of LUX-Lung 3. The subgroup analysis showed that the clinical effectiveness of afatinib differed according to family origin (Asian or non-Asian). This would also have an impact on the results of the mixed treatment comparison (which included the intention-to-treat population) which the ERG considered were not useful for decision-making. The ERG stated that the UK population is likely to be much closer in terms of characteristics and prognosis to the non-Asian subgroup than to the overall LUX-Lung 3 population who were predominantly of Asian origin.
3.23 The ERG questioned whether it was appropriate to include trials of EGFR mutation-positive populations with trials of unknown or mixed EGFR status populations in a single mixed treatment comparison. The ERG noted that there were differences in patient characteristics between studies of patients of EGFR mutation-positive NSCLC and those of unknown or mixed EGFR status in relation to the proportions of men, patients who had never smoked and patients with adenocarcinoma. The ERG also noted that the original mixed treatment comparison included patients with different histological types of NSCLC. The ERG concluded that the patient populations in the included trials were not sufficiently similar and therefore the results generated by the manufacturer's original mixed treatment comparison are not generalisable to a UK population.
3.24 During clarification the ERG requested additional sensitivity analyses on the mixed treatment comparison to assess the impact of local investigator assessments, updated overall survival data (if available), using only data from the population of patients with EGFR activating mutations for both progression-free survival and overall survival and excluding EURTAC trial data (because it included only European patients). The resulting hazard ratios (random-effects model) for the difference in median progression-free survival for afatinib compared with gefitinib were 0.50 (95% CI 0.02 to 10.72) when assessed by independent review and 0.48 (95% CI 0.03 to 9.57) when assessed by the trial investigator. The hazard ratio for the difference in median overall survival was 0.91 (95% CI 0.07 to 12.03) for afatinib compared with gefitinib. The ERG considered that the model should be populated with data from non-Asian patients to appraise the cost effectiveness of treatments for use in England; it is only appropriate to use data that have been generated from a non-Asian population of EGFR mutation-positive patients, whether in terms of primary clinical trials or supporting evidence for use in a simple indirect comparison or mixed treatment comparison. The ERG further highlighted an ongoing study (LUX-Lung 7) which directly compares afatinib and gefitinib in people with EGFR mutation-positive advanced NSCLC and is due to report in 2015.
3.25 The ERG disagreed with the approach taken by the manufacturer when fitting theoretical survival models to the LUX-Lung 3 data. The ERG did not consider that the Weibull models generated by the manufacturer for patients receiving afatinib or pemetrexed plus cisplatin accurately reflected the experience of LUX-Lung 3 patients, especially for progression-free survival which has an impact on the application of hazard ratios in the manufacturer's model. The ERG therefore considered that the progression-free survival results obtained from the Weibull model lacked credibility.
3.26 In view of the issues with the manufacturer's model, the ERG did not consider it appropriate to carry out an exploratory analysis using the manufacturer's model. The ERG specified that it was not possible to incorporate alternative survival projections into the model because it had been structured around the use of hazard ratios to generate survival estimates rather than using directly obtained estimates.
3.27 Because of the technical issues with the mixed treatment comparison, the ERG carried out an exploratory analysis to obtain an approximate estimate of the ICER for afatinib compared with combination chemotherapy. The results for the intention-to-treat population from LUX-Lung 3 showed that afatinib was associated with an ICER of £39,300 per QALY gained compared with pemetrexed plus cisplatin. The results for the non-Asian population showed that afatinib was associated with an ICER of £23,700 per QALY gained compared with pemetrexed plus cisplatin. The ERG concluded that the combination of patient access scheme pricing and use of data from the non-Asian subgroup of LUX-Lung 3 is likely to indicate that afatinib is cost effective compared with pemetrexed plus cisplatin in a predominantly white population of EGFR-positive patients.
3.28 The ERG also carried out a cost analysis of afatinib, erlotinib and gefitinib incorporating the patient access schemes which have been agreed by the Department of Health. Two separate analyses were undertaken, which differed with regards to the assumption of effectiveness. The first analysis assumed that patients experience the same overall survival hazard profile as experienced in the LUX-Lung 3 trial, but experience the individual progression-free survival hazard profile from the key clinical trial for each treatment (that is, IPASS for gefitinib, EURTAC for erlotinib and LUX-Lung 3 for afatinib). The second analysis assumed that patients experience both the same overall survival and progression-free survival hazard profiles as experienced in the LUX-Lung 3 trial, irrespective of treatment. Given the discounts of the patient access schemes for both afatinib and erlotinib are commercial in confidence, the results of the cost comparison cannot be presented here. The estimated cost per patient of gefitinib was £11,886 using the progression-free survival estimate from the IPASS trial and the same overall survival as afatinib, and £12,069 assuming the same overall survival and progression-free survival as afatinib.